Long (?) GC times in cc1plus

Brad Lucier lucier@math.purdue.edu
Wed Feb 9 08:35:00 GMT 2000


> From pfeifer@dbai.tuwien.ac.at  Tue Feb  8 13:11:50 2000
> Actually, we experience exactly the same behavior as Brad for hand-written
> C++ code to the extent that we have had to completely abandon using
> current CVS versions of GCC for testing a couple of months ago. :-(

I ran Gerald's C++ code on

Reading specs from /export/u10/egcs-profile/lib/gcc-lib/alphaev6-unknown-linux-gnu/2.96/specs
gcc version 2.96 20000127 (experimental)

with

options -ftemplate-depth-25 -O1 

and found the following statistics; I'm not familiar with statistics from
cc1plus, so it just seems to me that GC is taking a fair amount of time:

Each sample counts as 0.000976562 seconds.
  %   cumulative   self              self     total           
 time   seconds   seconds    calls  ms/call  ms/call  name    
  7.29      5.05     5.05 51168831     0.00     0.00  ggc_set_mark
  6.57      9.60     4.55   893689     0.01     0.01  ggc_mark_rtx_children
  5.45     13.38     3.78   482140     0.01     0.01  ggc_mark_tree_children
  4.61     16.57     3.20     1037     3.08     3.19  ggc_pop_context
  3.00     18.65     2.08  1673622     0.00     0.00  walk_tree
  2.65     20.49     1.83  7811247     0.00     0.00  ggc_alloc_obj
  2.16     21.99     1.50      848     1.77     4.62  init_alias_analysis
  1.99     23.36     1.38  2017888     0.00     0.00  fixup_var_refs_1
  1.46     24.37     1.01    11290     0.09     0.09  push_to_sequence
  1.17     25.18     0.81  4592375     0.00     0.00  note_stores
  1.03     25.89     0.71     7571     0.09     0.10  fixup_var_refs_insns

Everything else is below 1% of cpu time.  From the detailed report:

-----------------------------------------------
                0.01    1.39    1286/13195       expand_body <cycle 1> [47]
                0.03    3.59    3311/13195       rest_of_compilation <cycle 1> [5]
                0.09    9.33    8598/13195       cse_main <cycle 1> [10]
[7]     20.8    0.13   14.31   13195         ggc_collect [7]
                0.00   14.18      28/28          ggc_mark_roots [8]
                0.12    0.00  140397/280794      ggc_recalculate_in_use_p [168]
                0.01    0.00  140397/836433      xmalloc [386]
                0.00    0.00    2565/39346       set_page_table_entry [633]
                0.00    0.00      56/37953       get_run_time [893]
                0.00    0.00      28/28          release_pages [2502]
-----------------------------------------------

Brad


More information about the Gcc mailing list